Somewhere in the middle of a long grading cycle, an educator discovered that exhaustion had quietly replaced their metaphor-making mind with something mechanical and procedural—and that an AI, responding to flattened input, reflected the loss back by becoming unexpectedly poetic. The reversal was not a threat but a revelation: teaching is not a fixed identity but a cognitive stance that fatigue can erode without warning. In an era of anxious debate about what AI will take from educators, this moment offers a quieter and more instructive question—not what the machine might replace, but what it